Phenotypes of patients with extensive tooth wear—A novel approach using cluster analysis

Abstract

ObjectivesExplore a new approach to identify phenotypes of tooth wear (TW) patients using an unsupervised cluster analysis model, based on demographic, self-report, clinical, salivary and electromyographic (EMG) findings. MethodsData was collected for 34 variables from 125 patients, aged 17–65 years, with a TW index > grade 2. Demographic information and presumed risk factors for chemical and mechanical TW were collected. A 14-item stress scale was completed and salivary flow rates, pH and buffer capacity were measured. Sleep bruxism was assessed with a portable single channel EMG device. ResultsThe final cluster model comprised 16 variables and 103 patients and indicated two groups of TW patients; 61 participants in cluster A and 42 in cluster B. Cluster assignment was determined by several presumed mechanical risk factors and diseases affecting saliva. Cluster B had the highest percentage of sleep bruxism self-reports (A 1.6%, B 92.9%, p ≤ 0.001), awake bruxism self-reports (A 45.9%, B 85.7%, p ≤ 0.001), heavy sport exercises (A 1.6%, B 21.4%, p = 0.001); and highest percentage of diseases affecting saliva (A 13.1%, B 47.6%, p ≤ 0.001). A notable finding was the lack of significant differences between clusters in many other presumed risk factors for mechanical and chemical TW. ConclusionTW patients can be clustered in at least two groups with different phenotypic characteristics but also with a large degree of overlap. Based on this type of algorithm, tools for clinical application may be developed and underpin TW classification and treatment planning in the future.